Search

Although reinforcement learning (RL) can solve complex tasks after training, it's difficult to extend trained agents to target environments with perturbations. This lack of generalization ability has caused the dilemma of large-scale application of RL. Rich neural networks have brought huge improvements to deep learning, but these aren't adapted to RL and even bring negative effects. Therefore, RL algorithms have few choices of neural network, which greatly limits the representation and generalization ability of RL. To overcome these limitations, we propose a deep semi-dense compression network (DSCN) to improve RL generalization ability. First, we perform a structural extension on a general network model for RL. Then based on the information theory, we propose a semi-dense connection to enhance the neural information flow (NIF) of initial features. Finally, drawing on the ideas of course learning, we propose a channel compression approach to filter the redundant information of the deep network. In addition, we innovatively extend the experimental environment and evaluation metrics of the existing platform, which can fully evaluate the performance of DSCN. The experimental results show that our model achieves stable and significant generalization performance improvement. [ABSTRACT FROM AUTHOR]

As a powerful data modeling tool, Heterogeneous Information Network (HIN) has been successfully used in auxiliary information exploitation to boost recommendation performance. For HIN based recommendation, it is challenging to extract and fuse useful features of user preferences and item attributes under different semantic paths in HINs. Existing methods leverage a pre-defined fusion function to integrate different semantics for recommendation, which cannot characterize the complex nonlinear interactions between users and items. In this paper, we present a general framework named MNRec, short for Meta-path fusion based Neural Recommendation, to extract and fuse user and item embeddings under different meta-paths for recommendation. Under the framework, we propose an instantiation of MNRec with Multi-Layer Perceptron (MLP) structure. It consists of two major steps, i.e., meta-path based heterogeneous network embedding and deep learning based rating prediction. Concretely, appropriate meta-paths are first designed according to domain knowledge. Then the embeddings of users and items are obtained through a meta-path and commuting matrix based heterogeneous network embedding method. Finally, in light of the powerful nonlinear modeling capabilities of deep neural networks, the learned embeddings under different meta-paths are integrated into a two-pathway MLP structure for rating prediction. Experimental results on three real-world datasets demonstrate the superiority and effectiveness of MNRec compared with state-of-the-art baselines in rating prediction. [ABSTRACT FROM AUTHOR]

Deep learning has dominated the methodology of image inpainting in recent years. However, many of the existing methods still suffer from the defects of unnatural structure and blurry textures when filling large-area holes. To address this issue, we propose a novel coarse-to-fine residual inpainting framework: we first reconstruct the downsampled low-frequency coarse profile at a low computational cost, we then solely focus on generating high-frequency details, which will be added as residuals to the coarse profile, such that structure and texture details in composited result can be better preserved. Meanwhile, we notice that the existing attention mechanism used to build long-distance dependency is unreasonable: either does not adapt to free-form holes, or lacks filtering for invalid information sources, which could compromise feature map reconstruction. To solve this, we propose an improved Selective Free-Form Attention (SFFA) module, selectively allowing the information flow from the valid sources to hole area with free-form shape. We verify the incremental effectiveness of our contributions. We further show that our model outperforms recent state-of-the-art methods both qualitatively and quantitatively on benchmark datasets. [ABSTRACT FROM AUTHOR]

The assortments of global retailers are composed of hundreds of thousands of products linked by several types of relationships such as style compatibility, "bought together", "watched together", etc. Graphs are a natural representation for assortments, where products are nodes and relations are edges. Style compatibility relations are produced manually and do not cover the whole graph uniformly. We propose to use inductive learning to enhance a graph encoding style compatibility of a fashion assortment, leveraging rich node information comprising textual descriptions and visual data. Then, we show how the proposed graph enhancement substantially improves the performance on transductive tasks with a minor impact on graph sparsity. Although demonstrated in a challenging and novel industrial application case, the approach we propose is general enough to be applied to any node-level or edge-level prediction task in very sparse, large-scale networks. [ABSTRACT FROM AUTHOR]

Recommender systems are popular tools used in many applications, such as e-commerce, e-learning, and social networks to help users select their desired items. Collaborative filtering is a widely used recommendation technique that employs previous ratings of users to predict their future interests. Lack of sufficient ratings often reduces the performance of collaborative filtering recommendation methods. Additional side resources, such as trust relationships and tag information can be employed to enhance the recommendation accuracy. However, trust and tag data are often heavily sparse as users mainly provide insufficient information about these side resources. Moreover, such additional resources have generally large dimensions which result in increasing the computational complexity of recommendation models in calculating similarity values. To cope with these problems, a new recommendation model is proposed that utilizes deep neural networks to model the representation of trust relationships and tag information. To this end, a sparse autoencoder is used to extract latent features from user-user trust relationships and user-tag matrices. Then, the extracted latent features are utilized to calculate similarity values between users, which are then used to form the nearest neighbors of the target user and predict unseen items. The proposed method can tackle the data sparsity problem and reduce the computational complexity of recommender systems as the extracted latent features have smaller dimensions in comparison to the original data. Experimental results on two benchmark datasets reveal the effectiveness of the proposed method and show its outperformance over state-of-the-art recommender systems. [ABSTRACT FROM AUTHOR]

In this work, we examine the advantages of using multiple types of behaviours in recommendation systems. Intuitively, each user often takes some implicit actions (e.g., click) before making an explicit decision (e.g., purchase). Previous studies show that implicit and explicit feedback has different roles for a useful recommendation. However, these studies either exploit implicit and explicit behaviours separately or ignore the semantics of sequential interactions between users and items. In addition, we go from the hypothesis that a user's preferences at a time are combinations of long-term and short-term interests. In this paper, we propose some Deep Learning architectures. The first one is Implicit to Explicit (ITE) , to exploit users' interests through the sequence of their actions. The second and third ones are two versions of ITE with Bidirectional Encoder Representations from Transformers based (BERT-based) architecture called BERT-ITE and BERT-ITE-Si , which combine users' long- and short-term preferences without and with side information to enhance users' representations. The experimental results show that our models outperform previous state-of-the-art ones and also demonstrate our views on the effectiveness of exploiting the implicit to explicit order as well as combining long- and short-term preferences in three large-scale datasets. The source code of our paper is available at: https://github.com/tranquyenbk173/BERT_ITE. [ABSTRACT FROM AUTHOR]

Location-based Social Networks (LBSNs) enable users to socialize with friends and acquaintances by sharing their check-ins, opinions, photos, and reviews. A huge volume of data generated from LBSNs opens up a new avenue of research that gives birth to a new sub-field of recommendation systems, known as Point-of-Interest (POI) recommendation. A POI recommendation technique essentially exploits users' historical check-ins and other multi-modal information such as POI attributes and friendship network, to recommend the next set of POIs suitable for a user. A plethora of earlier works focus on traditional machine learning techniques that use hand-crafted features from the dataset. With the recent surge of deep learning research, we have witnessed a large variety of POI recommendation works utilizing different deep learning paradigms. These techniques largely vary in problem formulations, proposed techniques, used datasets and features, etc. To the best of our knowledge, this work is the first comprehensive survey of all major deep learning-based POI recommendation works. Our work categorizes and critically analyzes the recent POI recommendation works based on different deep learning paradigms and other relevant features. This review can be considered a cookbook for researchers or practitioners working in the area of POI recommendation. [ABSTRACT FROM AUTHOR]

[Display omitted] • To the best of our knowledge, our work is the first attempt to model search-and-go human behavior as a CTR-like binary classification problem in recommender system. • We propose an effective end-to-end deep learning model with a wide&deep architecture by utilizing LSTM attention and factorization-machine, to learn the sophisticated interactions between mobility data and search data. • We evaluate the framework on real-world dataset collected from Yahoo! Japan portal application, and demonstrate satisfactory performances on multiple search-and-go scenarios at a nation-wide level. • Our propsed model DeepSTIN achieved better performances comparing with the baseline models such as GBDT, DNN, Factorization-supported Neural Network [48], Product-based Neural Network [26], Wide&Deep [2], Deep & Cross Network [31], DeepFM [10], and XDeepFM [19]. • DeepSTIN improves the performance of DeepFM on four datasets by {6.4%, 7.9%, 11.8%, 9.3%} in AUC and {7.8%, 8.5%, 22.0%, 26.8%} in LogLoss. Every day, people are using search engines for different purposes such as research, shopping, or entertainment. Among the behaviors of search engine users, we are particularly interested in search-and-go behavior, which intuitively corresponds to a simple but challenging question, i.e., will users go where they search? Accurately estimating such behavior can be of great importance for Internet companies to recommend point-of-interest (POI), advertisement, and route, as well as for governments and public service operators like metro companies to conduct traffic monitoring, crowd management, and transportation scheduling. Therefore, in this study, we first collect search log data and GPS log data with linked and consistent user ID from Yahoo! Japan portal application installed in millions of smart-phones and tablets. Then we propose a framework including a complete data-processing procedure and an end-to-end deep learning model to predict whether a user will check-in the searched place or not. Specifically, as users' daily activities are considered to have high correlation with their travel, eating, and recreation decision in the future (i.e., go or not), Deep Spatial–Temporal Interaction Network (DeepSTIN) is elaborately designed to automatically learn the sophisticated spatiotemporal interactions between mobility data and search query data. Experimental results based on the standard metrics demonstrate that our proposed framework can achieve satisfactory performances on multiple real-world search scenarios. [ABSTRACT FROM AUTHOR]

Session-based recommendation, which aims to predict subsequent user actions based on anonymous sessions, plays a significant role in many online services. Existing methods construct each session as a digraph and then capture the rich transition relationship of items by using graph neural networks. However, their ability to obtain the user's static intentions is insufficient and they suffer from improper combinations of different user intentions. In this study, we propose a dynamic and static intentions integrated graph neural network (DSGNN) for session-based recommendation, in which the user's intentions captured by a digraph and an undigraph are comprehensively considered to enhance the recommendation performance. The prediction results from the dynamic and static intentions are combined using a lightweight gating network, which reduces the conflict between the two kinds of information. Furthermore, the weighted inner product is designed to alleviate the impact of the value of the item representation vectors. Extensive experiments on three real-world datasets show that the DSGNN outperforms other state-of-the-art methods and demonstrates that the components in our model are effective. [ABSTRACT FROM AUTHOR]

• We propose a fast, compact and parameter-efficient party-ignorant framework based on emotional recurrent unit. • We design generalized neural tensor block which is suitable for different structures, to perform context compositionality. • Experiments on three standard benchmarks indicate that our model outperforms the state of the art with fewer parameters. Sentiment analysis in conversations has gained increasing attention in recent years for the growing amount of applications it can serve, e.g., sentiment analysis, recommender systems, and human-robot interaction. The main difference between conversational sentiment analysis and single sentence sentiment analysis is the existence of context information that may influence the sentiment of an utterance in a dialogue. How to effectively encode contextual information in dialogues, however, remains a challenge. Existing approaches employ complicated deep learning structures to distinguish different parties in a conversation and then model the context information. In this paper, we propose a fast, compact and parameter-efficient party-ignorant framework named bidirectional emotional recurrent unit for conversational sentiment analysis. In our system, a generalized neural tensor block followed by a two-channel classifier is designed to perform context compositionality and sentiment classification, respectively. Extensive experiments on three standard datasets demonstrate that our model outperforms the state of the art in most cases. [ABSTRACT FROM AUTHOR]

• This study proposes novel approach for Quality of Service prediction for Cloud Service Recommendation. • A unique ensemble of Deep Learning models to learn user and service correlation is presented. • This approach combines neighbourhood based method and Neural Network models. • The proposed approach has better performance than other state-of-the-art approaches. Applications of Cloud Services are increasing day by day, and so is the difficulty of choosing the best-suited service for a customer. Quality of Service (QoS) parameters can be used for quality assurance and evaluation; further, a service can be recommended based on these QoS parameters' values. Recommendation systems are getting much attention lately. It has a crucial role in almost all the major commercial platforms and many improvements are being made to make the recommendations more precise and closer to the user's requirements. Conventional Machine Learning algorithms and statistical analysis methods, presently are not that efficient in learning the complex correlation between data elements. Lately, Deep Learning models have proven to be practical and precise in areas like natural language processing, image processing, data mining, & data interpretation. However, there are not many examples of complete Deep Learning applications for cloud service recommendation systems, though some works partially use Deep Learning. We propose the Ensemble of Deep Neural Networks (EDNN) method, which is of the hybrid type, i.e., the fusion of neighborhood-based and neural network model-based methods. The output obtained from both the models are combined using another different neural network model. Our approach for predicting QoS values is simple and different from previous works, and the results show that it outperforms other classical methods marginally. [ABSTRACT FROM AUTHOR]

As an effective deep representation learning technique for graph data, graph convolutional network (GCN) has recently been widely applied to obtain better embedding of vertex. The existing studies have successfully explored user-item interaction via GCN for recommendation task and proved its effectiveness. We argue that a significant limitation of these methods is that social relation, which has been proven to impose positive effects for recommendation in many loss optimization models, has received relatively little scrutiny in GCN. Thus, the resultant embeddings are insufficient to model the potential social propagation effect. In our work, we propose to obtain embeddings by using the neighborhood propagation mechanism on two coupled graphs, i.e. user-item interaction graph and social relation graph, which is capable of capturing the interplay between the user's item taste and user's friend relationship to integrate social effect into the embeddings. In particular, to address the challenge that different factors on neighborhood propagation process make different contributions for the embedding, we develop Attentional Social Recommendation system (ASR), a new social recommendation framework with hierarchical attention(i.e., neighbor-level and graph-level attention). This allows much flexibility for adaptively acquiring relative importance for different factors. Extensive experiment on three real-world datasets not only show the superior performance of our proposed model over the baselines, but also demonstrate the effectiveness of simultaneous neighborhood propagation on two graphs. [ABSTRACT FROM AUTHOR]

Recently, deep neural networks are widely applied in recommender systems for their effectiveness in capturing/modeling users' preferences. Especially, the attention mechanism in deep learning enables recommender systems to incorporate various features in an adaptive way. Specifically, as for the next item recommendation task, we have the following three observations: 1) users' sequential behavior records aggregate at time positions ("time-aggregation"), 2) users have personalized taste that is related to the "time-aggregation" phenomenon ("personalized time-aggregation"), and 3) users' short-term interests play an important role in the next item prediction/recommendation. In this paper, we propose a new T ime-aware L ong- and S hort-term A ttention N etwork (TLSAN) to address those observations mentioned above. Specifically, TLSAN consists of two main components. Firstly, TLSAN models "personalized time-aggregation" and learn user-specific temporal taste via trainable personalized time position embeddings with category-aware correlations in long-term behaviors. Secondly, long- and short-term feature-wise attention layers are proposed to effectively capture users' long- and short-term preferences for accurate recommendation. Especially, the attention mechanism enables TLSAN to utilize users' preferences in an adaptive way, and its usage in long- and short-term layers enhances TLSAN's ability of dealing with sparse interaction data. Extensive experiments are conducted on Amazon datasets from different fields (also with different size), and the results show that TLSAN outperforms state-of-the-art baselines in both capturing users' preferences and performing time-sensitive next-item recommendation. [ABSTRACT FROM AUTHOR]

The main objective of the recommender system is to determine user behavior or preference and suggest the items of interest to the user. We notice that topic information is one of the significant factors that enables a user to choose the next items. To mimic this behavior, we design a topic diversity discovering (TDD) model to capture user preferences on topics. Unlike the previous studies applied the topic information to recommender systems, which only considered the item topics. In this paper, we not only focus on the item topic information but also consider the user topic information on the recommender system. We develop a TDD model to learn the topic information from the item, and we then utilize the transfer learning to obtain the user topic information. We then integrate the TDD model with the Bayesian personalized ranking (BPR) model to build a novel TDD-BPR recommender system. We experiment on three well-known three datasets MovieLens-1 M, Pinterest, and TMDB dataset, which achieved 71.8%, 91.2%, and 96.1% in the Hit Ratio@10. Empirical evidence showed that TDD-BPR discover the potential topic information to understand the user's preferences and get better recommendation performance. [ABSTRACT FROM AUTHOR]

Existing food recommendation models have typically suggested foods or recipes to single users. However, in reality, users may be members of a group, family, or community, requiring food recommendation systems to support the whole group. Food recommendations to groups are a more challenging task than food recommendations to individuals, as each person's preferences in the group should be addressed before giving the recommendations. Suggesting healthy food is also important in a food recommendation system, given that unhealthy diets can lead to different diseases. To address these challenges, a new healthy group food recommendation system based on deep social community detection and user popularity is developed in this study. To this end, an innovative deep community detection approach based on feature learning and deep neural networks is developed using the calculated time-aware user similarity measure. In addition, a health-aware rate prediction measurement, which considers both group preferences and health factors, is developed. Different experiments are designed on two real-food social networks to specify the efficiency of the suggested model, and the results indicate that it enhanced the single-user and group satisfaction metrics. • Developing a novel healthy food recommendation to user groups. • Utilizing deep community detection and popularity measure in recommendation. • Considering user ratings, food ingredients and nutrition in the recommendation. • Results obtained from two real datasets demonstrated the efficacy of our model. [ABSTRACT FROM AUTHOR]

With the development of e-commerce platforms, user reviews have become a vital source of information to address the sparsity problems for enhancing the predictive performance of the recommendation systems (RSs). However, the traditional methods of the RSs used to model user/item latent features based on static vectors in an independent manner without considering the dynamic nature of the user-item interactions which potentially affect the accuracy of the recommendation process. Thus, this paper proposes a RS model that exploits neural attention techniques to learn user/item representations by jointly considering the fine-grained semantic information for the user-item pairs. The proposed model utilizes both review-based and interaction-specific features for the user/item reviews to learn heterogeneous user/item representations. First, a BiLSTM sequence encoder is used to learn the contextual information of words, and a Co-attention network is then designed to jointly capture the most relevant semantic information of reviews for the user-item pair. To better capture user/item latent factors comprehensively, interaction-specific features based on the rating scores are further integrated with the review-specific latent features via a shared hidden layer. Finally, an attentive factorization machine (FM) is then applied on the shared hidden layer of the integrated user/item features for the final prediction. We carry out a series of experiments using real-world datasets and the results demonstrate that our proposed method is better than the baseline approaches in terms of both rating prediction and ranking performance. [ABSTRACT FROM AUTHOR]

Cross-domain recommendation is becoming increasingly popular recently. Existing cross-domain recommendation often assumes that, a sufficient set of bridged users across domains is given in advance which disregards the scenario with insufficient bridged users. In this paper, we propose a novel Trust-aware Latent Space Mapping approach for Cross-domain Recommendation, called TLSM-CDR. This represents one of the first attempts to address the challenge of insufficient bridged users from the perspective of users' trust relationships to facilitate user sharing cross-domain recommendation. First, our model employs the Probabilistic Matrix Factorization (PMF) to generate user and item matrices. Then, Deep Neural Network (DNN) and graph Laplacian are seamlessly incorporated into our trust-aware non-linear mapping function to capture the latent space relationships between both bridged and non-bridged users. Finally, we predict the optimized users' rating matrix in the target domain. Extensive experiments conducted on two real-world datasets demonstrate that, our TLSM-CDR model significantly outperforms several state-of-the-art methods. [ABSTRACT FROM AUTHOR]

• We propose MetaGen: which can able to provide solutions for two tasks such as peer review prediction, meta-review generation. • Integrated framework of convolution layer, Bi-LSTM model, and attention mechanism to predict the final decision. • Provide a concise meta-review which maximizes information coverage, coherence, readability and also reduces redundancy. • Experiments on PeerRead dataset shows that MetaGen outperforms state-of-the-art methods. Peer reviews form an essential part of scientific communications. Research papers and proposals are reviewed by several peers before they are finally accepted or rejected for publication and funding, respectively. With the steady increase in the number of research domains, scholarly venues (journal and/or conference), researchers, and papers, managing the peer review process is becoming a daunting task. Application of recommender systems to assist peer reviewing is, therefore, being explored and becoming an emerging research area. In this paper, we present a deep learning network based Meta-Review Generation considering peer review prediction of the scholarly article (MRGen). MRGen is able to provide solutions for: (i) Peer review prediction (Task 1) and (ii) Meta-review generation (Task 2). First, the system takes the peer reviews as input and produces a draft meta-review. Then it employs an integrated framework of convolution layer, long short-term memory (LSTM) model, Bi-LSTM model, and attention mechanism to predict the final decision (accept/reject) of the scholarly article. Based on the final decision, the proposed model MRGen incorporates Pointer Generator Network-based abstractive summarization to generate the final meta-review. The focus of our approach is to give a concise meta-review that maximizes information coverage, coherence, readability and also reduces redundancy. Extensive experiments conducted on the PeerRead dataset demonstrate good consistency between the recommended decisions and original decisions. We also compare the performance of MRGen with some of the existing state-of-the-art multi-document summarization methods. The system also outperforms a few existing models based on accuracy, Rouge scores, readability, non-redundancy, and cohesion. [ABSTRACT FROM AUTHOR]

Reinforcement learning (RL) techniques have recently been introduced to recommender systems. Most existing research works focus on designing policy and learning algorithms of the recommender agent but seldom care about the top-aware issue , i.e., the performance on the top positions is not satisfying, which is crucial for real applications. To address the drawback, we propose a Supervised deep Reinforcement learning Recommendation framework named as SRR. Within this framework, we utilize a supervised learning (SL) model to partially guide the learning of recommendation policy, where the supervision signal and RL signal are jointly employed and updated in a complementary fashion. We empirically find that suitable weak supervision helps to balance the immediate reward and the long-term reward, which nicely addresses the top-aware issue in RL based recommendation. Moreover, we perform a further investigation on how different supervision signals impact on recommendation policy. Extensive experiments are carried out on two real-world datasets under both the offline and simulated online evaluation settings, and the results demonstrate that the proposed methods indeed resolve the top-aware issue without much performance sacrifice in the long-run, compared with the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

• Recurrent convolutional networks are effective for session-based recommendation. • A unified and flexible network architecture with CNN and RNN is proposed. • Combining CNN with RNN allows capturing important features of session sequences. • The proposed combined model improves recommendation accuracy. Traditional recommender systems help users find items of interest by modeling long-term user profiles, which consist of items the users interacted with in the past. In many real-life applications, however, informative user profiles are not available. Instead, the system recommends by relying on the current user activities within an ongoing session, leading to the emergence of session-based recommendation methods. Among techniques used in such situations, recurrent neural networks (RNNs) present a natural choice thanks to their ability to model the order of session events and capture long-term dependencies. Recently, methods based on convolutional neural networks (CNNs) have also shown their potential in modeling session data, especially in extracting complex local patterns that are predictive of the user target. In this work, we propose a recurrent convolutional architecture that takes the advantages of both complex local features extracted by CNNs and long-term dependencies learned by RNNs from session sequences. Our model has two main layers: the lower layer consists of convolutional filters applied over consecutive session event embeddings, and the upper layer is a gated recurrent unit (GRU) RNN that takes as input the CNN's output. This hybrid approach provides a flexible and unified network architecture for modeling various important features of session sequences. Experiments conducted on three benchmark datasets demonstrate the superiority of the proposed model over pure RNNs and CNNs models as well as state-of-the-art session-based recommendation methods. [ABSTRACT FROM AUTHOR]

Rating prediction has long been a popular research topic in the field of recommendation systems. Latent factor models, particularly matrix factorization (MF) methods, constitute the most prevalent techniques for rating prediction. However, MF-based methods suffer from the problems of data sparsity and lack of explicability. In this paper, we propose a novel model in which ratings and topic-level review information are integrated into a deep neural framework to address these problems. Specifically, we designed a topic alignment operation and a topic attention mechanism to reflect the user's preferences and an item's properties in terms of a topic in the reviews. We present a neural prediction layer that extends user and item representations by including both the latent factors from ratings and the textual information from reviews. The results of extensive experiments on four real-world datasets demonstrate that our proposed method consistently outperforms the state-of-the-art recommendation approaches that follow this direction in terms of rating prediction. Furthermore, our model can categorize representative reviews of users/items and group reviews into topics for the users' reference. [ABSTRACT FROM AUTHOR]

User-generated product reviews contain a lot of valuable information including users' opinions on products and product features that is not fully exploited by the current recommendation models. Similarly, the metadata information related to the products, about the reviews and about the users who have written the reviews has rarely been exploited for recommender systems. These heterogeneous information sources have the potential to alleviate the cold start and sparsity problems and improve the quality of recommendations. In this paper, we present a joint deep recommendation model (JDRM) that consists of two parallel neural networks, learning lower-order feature interactions of users and items separately and higher-order feature interactions jointly using a shared last layer. Each of the networks is further composed of two sub-networks. One of the sub-networks focus on exploiting product reviews (of user/item) and the other sub-network learns user preferences/items properties leveraging metadata information along with the ratings. The learned latent features in each network are concatenated, thus producing the user and item latent feature vectors. We combine the two networks by introducing a shared layer on the top, which is a dense fully connected layer used to learn higher level latent features obtained from the two networks and produces final ratings. Extensive experiments on real-world datasets demonstrate that JDRM significantly outperforms state of the art recommendation models. [ABSTRACT FROM AUTHOR]

Knowledge Graphs (KGs) have been applied to various application scenarios including Web searching, Q&A, recommendation system, natural language processing and so on. However, the vast majority of Knowledge Bases (KBs) are incomplete, necessitating a demand for KB completion (KBC). Methods of KBC used in the mainstream current knowledge base include the latent factor model, the random walk model and recent popular methods based on reinforcement learning, which performs well in their respective areas of expertise. Recurrent neural network (RNN) and its variants model temporal data by remembering information for long periods, however, whether they also have the ability to use the information they have already remembered to achieve complex reasoning in the knowledge graph. In this paper, we produce a novel framework (ALSTM) based on the Attention mechanism and Long Short-Term Memory (LSTM), which associates structure learning with parameter learning of first-order logical rules in an end-to-end differentiable neural networks model. In this framework, we designed a memory system and employed a multi-head dot product attention (MHDPA) to interact and update the memories embedded in the memory system for reasoning purposes. This is also consistent with the process of human cognition and reasoning, looking for enlightenment for the future in historical memory. In addition, we explored the use of inductive bias in deep learning to facilitate learning of entities, relations, and rules. Experiments establish the efficiency and effectiveness of our model and show that our method achieves better performance in tasks which include fact prediction and link prediction than baseline models on several benchmark datasets such as WN18RR, FB15K-237 and NELL-995. [ABSTRACT FROM AUTHOR]

• We extend a previous work on multiclass preference context (MPC) to both user-oriented MPC and itemoriented MPC. • We propose a unified framework called matrix factorization with heterogeneous multiclass preference context (MFHMPC). • We have MF with dual MPC (MF-DMPC) for concurrent structure and MF with pipelined MPC (MF-PMPC) for sequential structure. • We conduct extensive comparative empirical studies and show the effectiveness of our MF-DMPC and MF-PMPC. The spreading use of the Internet and big data technology has spawned the need for recommendation systems. However, to alleviate public anxiety about privacy, this paper advocates making recommendation with internal context, which refers to the implicit context hidden beneath the rating matrix only. Inspired by a recent work that embeds neighborhood information among users (as represented by multiclass preference context, MPC) into a factorization-based method, we extend it to both user-oriented and item-oriented, and put forward both user-oriented MPC and item-oriented MPC into a generic factorization framework called matrix factorization with heterogeneous MPC (MF-HMPC). In particular, we derive two specific recommendation methods from MF-HMPC with different structures, including MF with dual MPC (MF-DMPC) and MF with pipelined MPC (MF-PMPC), which are corresponding to concurrent structure and sequential structure, respectively. Extensive empirical studies on four public datasets show that our two forms of MF-HMPC outperform the referenced state-of-the-art MF methods as well as two representative deep learning methods. Moreover, we also discover some interesting facts about how these two kinds of internal contextual information complement each other. The main advantage of our methods is that they manage to strike a good balance between user-oriented neighborhood information and item-oriented neighborhood information. [ABSTRACT FROM AUTHOR]

Recent work shows that incorporating tag information to recommender systems is promising for improving the recommendation accuracy in social systems. However, existing approaches suffer from less reasonable assignment of tag weights when constructing the user profiles and item characteristics in real-world scenarios, resulting in decreased accuracy in making recommendations. The above issue is specifically summarized into two aspects: 1) the weight of a target item is mainly determined by number of one certain type of tags, and 2) users place equal focus on the same tag for different items. To tackle these problems, we propose a novel model named TNAM, a T ag-aware N eural A ttention M odel, which accurately captures users' special attention to tags of items. In the proposed model, we design a tag-based neural attention network by extracting potential tag information to overcome the difficulty of assigning tag weights for personalized users. We combine user-item interactions with tag information to map sparse data to dense vectors in higher-order space. In this way, TNAM acquires more interrelations between users and items to make recommendations more accurate. Extensive experiments of our model on three publicly implicit feedback datasets reveal significant improvements on the metrics of HR and NDCG in Top-N recommendation tasks over several state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

• We define the historical users, which is calculated to capture complex interactions among users and items • We propose DeepUCF framework, which integrates a Deep+Shallow pattern to capture relationships between users. • We design an attention network to distinguish the importance of historical users, which we call DeepUCF+a. The user-based collaborative filtering (UCF) model has been widely used in industry for recommender systems. UCF predicts a user's interest in an item based on rating information from similar user profiles. A neural network UCF model can learn effectively the high-order relations between users and items, but it cannot distinguish the importance of users in learning process. To mine the complex relationships between users and items, we incorporate a Deep+Shadow pattern to improve learning features effectively, namely as DeepUCF. Firstly, we define historical users, that is, users from the historical data of interactions with an item. The target user and historical users are calculated to capture complex process of user's interacted item. Secondly, we integrate a shallow linear model to effectively solve single pair interaction problems. Finally, DeepUCF construct of a pair of user relations (interactive users with a history of items) for the input, and joint linear and nonlinear models to build relationships between users. More importantly, DeepUCF+a add an attention network to distinguish the historical user importance of items, which make DeepUCF more expressive. Experiments on real datasets show that DeepUCF and DeepUCF+a can effectively capture users' complex high-order relationships, and achieve better performance. [ABSTRACT FROM AUTHOR]

Nowadays, people prefer to read news articles from online sources worldwide due to their easiness and availability. For the last few years, online searching for required information or content has been replaced by item recommendation, and news recommendation is also not an exception. For news recommendations, News Recommender System (NRS) helps the users to find the appropriate and pertinent content, alleviate the problem of information overload, and propose news that would be of interest to news readers. NRS also assists different users all around the world in this regard by recommending the most recent news articles based on their interests and past preferences. Many techniques such as traditional, Deep Learning (DL), and hybrid have been proposed to solve the NRS challenges and issues. DL techniques are considered one of the best techniques and have been successfully applied in various fields such as Natural Language Processing (NLP) and Computer Vision (CV). This survey article provides a detailed analysis of DL models-based techniques to build NRS. In this regard, firstly, a comprehensive comparison is provided between published survey articles on NRS and this research work. Secondly, it discusses the background of recommendation systems and their techniques. Furthermore, NRS is explored along with its current research challenges. Then background knowledge of DL and its methods have been discussed along with the analysis of year-wise published relevant articles having DL as the applied technique. The survey also presents widely used datasets and performance evaluation metrics used in the relevant literature. Finally, a detailed discussion provides several future directions and open research challenges for the researchers to consider DL applications in NRS. [ABSTRACT FROM AUTHOR]

Abstract Traditional recommendation methods based on matrix factorization techniques have yielded immense success because of their good scalability. However, they still face the problem of data sparsity, which may lead to a reduction in recommendation performance. As it is hard to learn good latent features in the sparse user-item rating matrix. In recent years, deep learning is very appealing in learning effective representations. Its non-linear characteristics just remedy the shortcomings of matrix factorization. In this paper, a novel method deep variational matrix factorization recommendation (DVMF) is proposed for large scale sparse dataset. DVMF is based on latent factors to predict the ratings. The latent features of the users and items are respectively obtained through a deep nonlinear structure. Based on the latent factors and combined with matrix factorization method, the paper presents algorithm optimization method of DVMF. The experiments on three real-world datasets from different domains show that DVMF is able to provide higher accuracy than recommendation algorithms based on matrix factorization or deep learning individually on large scale sparse dataset. [ABSTRACT FROM AUTHOR]

Abstract Recently, recommendations accompanied by review text have become increasingly common. Some studies have incorporated user and product information to generate review representations; however, the corresponding methods generally suffer from high model complexity and only consider textual review information at the word level rather than the semantic level. To address the above issues, we present a hierarchical attention model called HAUP based on user and product reviews. The model jointly learns user and product information from ratings and user review text in recommendation. First, a hierarchical bidirectional gated recurrent unit (Bi-GRU) is constructed that includes word and sentence-level review information. The Bi-GRU structure can handle long-range dependency in the review text. Then, attention mechanisms are applied at the word and sentence levels to identify the most important content when constructing the review document representation. Finally, the generated latent user and product representations are merged into the same vector space for estimating ratings. Experiments conducted based on two real-world data sets demonstrate that the proposed architecture outperforms existing state-of-the-art models. A visualization of the different attention layers demonstrates that the proposed model selects important words and sentences in review documents. [ABSTRACT FROM AUTHOR]

Feature engineering is a classical problem in recommender systems, and feature interactions is one of the most important parts of feature engineering. Factorization based models are widely used for explicit feature interactions. However, most current works utilize separate features to model cross features. Such a pattern limits the significance of cross features, since realistic recommendation scenarios are rich in associations between features. In this paper, we classify the basic feature interactions into sum-interaction and product-interaction, and improve the current general strategy of explicit feature interactions. Based on these theoretical studies, we propose a novel explicit feature interactions model Attentional Aggregative Interaction Network (AAIN), which models higher-order features using a cyclic explicit module. Specifically, we introduce attention mechanism for the reorganization of separate features, followed by product-interaction and higher-order features' compression and output. The model is efficient since: 1) AAIN automatically learns high-order feature interactions and filters them with different weights. 2) AAIN optimizes the interaction between features into the interaction between feature groups, which allows for other relevant information to be considered when performing interactions. Furthermore, we integrate AAIN model with the classical deep neural network (DNN) model into a new model Deep Attentional Aggregative Interaction Network (DAAIN). Experiments on real-world datasets show that our models achieve state-of-the-art results. [ABSTRACT FROM AUTHOR]

Neighborhood-based collaborative filtering is a method of high significance among recommender systems, with advantages of simplicity and justifiability. However, recently it is receiving less popularity due to its low prediction accuracy in contrast with model-based collaborative filtering systems, but model-based methods also suffer from a drawback worthy of attention that is they cannot effectively explain the reason behind their estimation. In order to develop a system with both high accuracy and justifiability, we propose a novel neighborhood-based collaborative filtering method inspired by the natural mechanism of attention. Our method can adaptively find neighborhood items to the prediction in user history without any pre-defined function with respect item correlations. Then the estimation are made based on these relationships. Experiments on several benchmarks are carried out to verify the performance of the proposed method, and the result shows that our method beats all previous state-of-the-art methods on MovieLens 10M and Netflix in addition to being able to justify the prediction obtained. [ABSTRACT FROM AUTHOR]